Several studies revealed that predicting one's memory during learning has reactive effects on memory performance: making judgments of learning (JOLs) improved cued-recall performance for related word pairs and recognition memory for single words but impaired or did not impact cued recall for unrelated word pairs. Very little is known, however, about the generalizability of JOL reactivity to educationally relevant study materials and test conditions. The three experiments reported here examined whether making JOLs affects performance in multiple-choice and short-answer tests on general knowledge facts of varying difficulty. Neither experiment revealed reliable effects of making JOLs on performance, and the same was found in a continuously cumulating meta-analysis that included data from all experiments. These results indicate that soliciting metamemory predictions does not affect test performance for general knowledge facts. Further research will be needed to examine whether this holds for educationally relevant materials and tests in general.